Imagine trying to send a giant high-definition movie over a slow internet connection; it usually crashes or takes forever. This project creates a smart 'shorthand' for satellite and weather data, shrinking massive files by up to 1000 times without losing the important details. It allows computers to quickly find patterns across different data sources without needing to move petabytes of information around.
Transferring petabytes of satellite and weather data between archives is too slow and expensive, making many real-time geospatial applications economically impossible.
An openEO API extension for neural compression, the NeuCo-Bench benchmarking tool, and four application prototypes for maritime, biomass, climate, and agriculture.
If you are a precision farming service provider dealing with massive satellite imagery for crop stress and yield detection — this project developed AI embeddings that compress data up to 1000-fold. This makes it economically viable to analyze continental-scale fields in near-real-time.
If you are a shipping monitor dealing with the high cost of transferring petabytes of SAR and AIS data — this project developed a federated deployment pipeline for vessel detection. This reduces 'data gravity' and lowers the computational demand for maritime awareness.
If you are a climate risk firm dealing with fragmented weather and biomass data archives — this project developed an openEO API extension for on-demand neural compression. This allows you to fuse diverse geo-information without downloading massive raw datasets.
Based on available project data, specific pricing or licensing costs are not provided, as the project focuses on open-sourcing and standardization.
Yes, the project is specifically designed for petabyte-scale spatio-temporal data and targets a 1000-fold reduction in data size to enable continental-scale assessments.
The project emphasizes open-sourcing, having already released the NeuCo-Bench benchmarking tool and SSL4EO-S12-downstream datasets on HuggingFace.
It provides an extension to the openEO API, allowing for seamless integration of neural compression endpoints across different data centers.
The project period runs from 2024-01-01 to 2026-12-31, with prototypes already co-developed in four key domains.
The consortium is well-balanced for commercialization, featuring a 40% industry ratio with 4 industrial partners and 1 SME. With 10 partners across 6 countries (including the UK, NL, and DE), the group combines high-performance computing (HPC) expertise with academic research and practical industry application, ensuring the technology is grounded in real-world operational needs.
Contact MARTEL INNOVATE BV in the Netherlands for API integration details.
Request access to the NeuCo-Bench framework for your geospatial data pipeline.
